Add conditional fields support

[ojab]

See #81
One omission: methods should be declared before field declaration using them, i. e.

class UserBlueprinter < Blueprinter::Base
  # field :first_name, if: :my_method here will fail
  def my_method(user, options)
    # some code
  end

  field :first_name, if: :my_method
end

because methods are not available yet at field declaration time and I'm resolving Symbol to Method there. Is it worth fixing or I should just add a note to README?

Other one: AFAICS collection serialization is not tested at the moment, so I also skipped it.

[philipqnguyen]

Thanks for making this PR. This will be a great addition. What do you think of the suggestions?

[philipqnguyen]

There might be a way to resolve the conditional method at render execution time rather than at field declaration time, allowing conditional methods to be defined anywhere in the blueprint. Here's the idea, let me know what you think =)

What if you move this chunk of code into the skip? method you made in Field?

[philipqnguyen]

We would need to pass self here. field.skip?(object, local_options, self)

[philipqnguyen]

Argument list would be altered to look like: skip?(object, local_options, blueprint)

Then you can place your above code here in this method:

    %i[if unless].each do |condition|
      if options[condition].is_a?(Symbol)
        options[condition] = blueprint.method(options.fetch(condition))
      end
    end

[philipqnguyen]

And as for collection testing... yes we need to get that in there. It would be in a separate PR though, so I wouldn't worry about it here.

[ojab]

Passing blueprint to Field at initialization time looks more clear to me.
Pushed that version, what do you think?

[philipqnguyen]

Passing self during field initialization makes sense! Really like this approach.

I have a few requests regarding the organization of the specs.

[philipqnguyen]

The current location of this context block makes it look like :unless depends on first having :if passed to field. However, both :if and :unless should be at the same level of context right?

I'm thinking of something like:

    Inside Rails project
      Given blueprint has ::field with a conditional argument
        Given the conditional is :if
           ... other contexts / specs
        Given the conditional is :unless
           ... other contexts / spces
    Outside Rails project
      ... repeat from above ...

[ojab]

it actually is: it's specs for both :if & :unless passed. Added specs for :unless only for clarity and fixed examples descriptions, now it looks like:

    Inside Rails project
      Given blueprint has ::field with a conditional argument
        Given the conditional is :if proc returning true
          and no :unless conditional
            serializes the conditional field
          and :unless conditional is proc returning true
            does not serialize the conditional field
          …
        Given the conditional is :unless proc returning true and no :if conditional
          does not serialize the conditional field
        Given the conditional is :if proc returning false
          and no :unless conditional
            does not serialize the conditional field
          and :unless conditional is proc returning true
          …

[philipqnguyen]

The current rspec printout out for these it blocks are something like:

and :unless is proc returning false
  should eq "{\"first_name\":\"Meg\",\"id\":15}"

or

and :unless is proc returning true
  should eq "{\"id\":16}"`

When reading the rspec printout, we don't actually care whether first_name is there or not, as that is just the example. What we care is whether the field with the conditional is there or not.

I think a description for the it block should be there to clarify the printout.
it ('does not serialize the field') { should eq(result_without_first_name) }
or
it ('serializes the field') { should eq(result_with_first_name) }

which gives a printout of something like:

and :unless is proc returning true
  serialize the field

or

and :unless is proc returning false
  does not serialize the field

This comment applies to the other it blocks in this PR as well.

[philipqnguyen]

released in 0.6.0